This paper focuses on detecting anomalies in Internet backbone traffic. To monitor traffic on a scale of several terabits per second, we need to divide the time series data of a traffic volume into many slices. Therefore, we need to monitor a lot of traffic data. However, adjusting an appropriate threshold for each traffic time series data individually is difficult. To solve this problem, we propose an anomaly-detection algorithm that does not need parameters to be set for each time series data. This algorithm operates acc-urately with low computational complexity. A side-by-side test demonstrated that the accuracy of the algorithm was higher than that of the conventional method. Moreover, the necessary learning period of the algorithm was shorter than that of the conventional method.